Dear Paul,
Having looked at your images I think the problem is that these subjects are quite extreme, in terms of atrophy and/or asymmetry. Therefore the default linear registration isn't going to be enough. It would be worth trying the non-linear registration option in this case. Once you have a non-linear registration of your image to the standard template (using flirt and then fnirt) you can use the -wfnirt option in run_first and see if that improves things. I am not sure if this will work for all structures though, as there is very little left of some structures in a couple of your images, and it might be difficult to get anything to work well for them.
All the best,
Mark
On 29 Oct 2013, at 07:50, Paul Chou <[log in to unmask]> wrote:
> Dear FSL developers and experts
>
> Thanks for your great effort for developing this wonderful tool. Currently I am focusing on subcortical segmentation of large aging cohort study (around 500-700 subjects). I found the FIRST works great for most participant's T1 image, but have some mis-segmentation problem on some specific dataset (5 of 460 subjects didn't performed well using FIRST). I also found the mis-segmentation area mainly located on caudate which probably due to the large ventricle. Would you have a look on these datasets and give me some advices and solutions to solve this mis-segmentation problem, thank you.
>
> The dataset can be downloaded at : https://www.dropbox.com/sh/5bdwli9v7lx43x1/pioCNAf52z
>
> Best
>
> Paul
|